Seismic noise interferometry is an exciting technique for studying volcanoes because it provides a continuous measurement of relative seismic velocity (dv/v), which is sensitive to magmatic processes that affect the physical properties of the surrounding crust. However, understanding the causal mechanisms controlling dv/v can be difficult and seasonal environmental stresses can obscure volcanic signals.

I use volcanic tremor (∼ 0.3 to 1.0 Hz) at Kīlauea summit, Hawai’i, as a passive source for interferometry. Noise cross-correlation functions (CCFs), calculated between 230 station pairs, have coherent and temporally consistent coda wave signals, from which small changes in arrival times of the highly scattered energy are measured. The resulting time series of relative seismic velocity shows a remarkable correlation with the radial tilt record measured at the summit, consistently correlating on a time scale of days to weeks for almost the entire study period (June 2011 to November 2015). As the summit continually deforms in deflation-inflation events, the velocity decreases and increases, respectively. Modelling of strain at Kīlauea suggests that, during inflation of the shallow magma reservoir (1 to 2 km below the surface), most of the edifice is dominated by compression - hence closing cracks and producing higher velocities - and vice versa.

In central Iceland, I measure dv/v from ambient seismic noise recorded at 51 stations from 2008 to 2018, across a range of frequency bands between 0.1 and 16 Hz. A clear annual cycle in dv/v is modelled as resulting from elastic and poro-elastic responses to changing snow depth, groundwater level and atmospheric pressure. This seasonal pattern ultimately needs to be accounted for if dv/v is to be used as a monitoring tool in Iceland. Furthermore, I observe a linear correlation between changes in dv/v and volumetric strain at stations in regions of both compression and dilatation associated with the 2014 Bárðarbunga-Holuhraun dyke intrusion. The intense seismicity associated with the subsequent rifting event alters the CCFs, making measurements of dv/v unreliable during the 6-month eruption. However, volcanic tremor can be located by measuring the arrival times of the main peak in the CCFs of pairs across the network. I locate tremor under Vatnajökull glacier, associated with small sub-glacial eruptions, and at the main vents in the Holuhraun lava field.

I comprehensively explain variations in dv/v arising from diverse crustal stresses, including those from magmatic pressurisation and intrusions, environmental loads and varying pore-pressure. dv/v can therefore complement other geophysical measurements, including those of surface deformation and microseismicity, in order to understand the dynamics of a volcanic system.